Accelerating Garfield++ with CUDA Enables Detailed Gaseous Detector Simulations Involving Thousands of Electrons

Comprehensive detector simulations are essential for modern particle physics, and Garfield++ serves as a widely used tool throughout the entire experimental process. To address the increasing computational demands of simulating advanced micro-pattern gaseous detectors, T. Neep, K. Nikolopoulos, and M. Slater developed a method to significantly accelerate one of Garfield++’s most intensive algorithms. The team achieved this by adapting the software to utilise the power of graphics processing units through the CUDA framework, integrating the changes directly into the Garfield++ codebase with minimal disruption for users. This advancement enables more efficient and detailed simulations, particularly for high-gain avalanches involving large numbers of electrons, ultimately allowing researchers to design and calibrate detectors with greater precision.

The operation and calibration of micro-pattern gaseous detectors require computationally intensive microscopic avalanche simulations. This work describes the acceleration of Garfield++’s most demanding algorithm, AvalancheMicroscopic, by porting it to graphics processing units using NVIDIA’s CUDA framework. The modifications are integrated into the Garfield++ codebase and are accessible to end users with only minor adjustments to their existing code. Benchmark results demonstrate substantial speed-up, especially for high-gain avalanches involving thousands of electrons, thereby enabling more efficient and detailed detector simulations.

GPU Acceleration of Avalanche Simulations in Garfield++

Scientists have successfully integrated graphics processing unit (GPU) support into Garfield++, a widely used software toolkit for simulating gaseous detectors, significantly accelerating computationally intensive microscopic avalanche simulations. The team focused on optimizing AvalancheMicroscopic, a detailed algorithm modelling charge transport at the collision level, which previously presented a substantial computational bottleneck. This work delivers a pathway for existing Garfield++ users to leverage the power of GPUs with minimal code modification, maintaining compatibility with a well-established and thoroughly tested codebase. The research demonstrates substantial speed-up in simulating high-gain avalanches, involving thousands of electrons, thereby enabling more efficient and detailed detector modelling.

Implementation involved adapting data structures and algorithms to execute on GPUs using NVIDIA’s CUDA framework, while prioritizing minimal disruption to the existing Garfield++ code. Validation studies confirm the accuracy of results obtained using the GPU-accelerated AvalancheMicroscopic algorithm, ensuring consistency with CPU-based simulations. This breakthrough enables researchers to explore a wider range of detector designs and operating conditions, ultimately improving the sensitivity and performance of gaseous detectors used in diverse applications. The modified Garfield++ codebase, incorporating GPU support, has been publicly available since June 2024, providing the detector simulation community with a powerful new tool for advancing their research.

GPU Acceleration Boosts Avalanche Simulation Speed

This work successfully integrates graphics processing unit (GPU) acceleration into the widely used Garfield++ software toolkit, specifically targeting the computationally demanding AvalancheMicroscopic algorithm. By leveraging NVIDIA’s CUDA framework, the team achieved substantial speed-up in simulating microscopic avalanche processes within gaseous detectors, a crucial aspect of modelling modern micro-pattern detector technologies. This advancement allows for more efficient and detailed simulations, enabling researchers to better understand and optimise detector performance throughout the experimental lifecycle. The implementation prioritised minimal disruption to existing Garfield++ code, ensuring a smooth transition for current users and facilitating the adoption of hardware acceleration without requiring extensive code modifications. The results demonstrate a significant performance gain for high-gain avalanches involving large numbers of electrons, and the authors suggest that exploring alternative GPU programming models represents a potential avenue for future research. This work provides a valuable step towards more realistic and computationally feasible simulations of gaseous detectors, ultimately contributing to advancements in particle detection technologies.